摘要
遥感影像去噪对于影像后续的使用和研究具有重要意义。高斯噪声与椒盐噪声是影像中常见的噪声,目前的去噪算法对于这类混合噪声普遍存在去噪效果不佳、去噪后影像边缘模糊等缺点。针对以上问题,提出了一种遥感影像混合噪声二阶去除方法。该方法第一阶段是在DnCNN网络框架的基础上引入扩张卷积来增加网络的感受野,便于在遥感影像中提取更多的特征信息;同时在深卷积层后引入DropoutLayer层构建降噪模型,以防止网络出现过拟合,简化训练难度,然后使用该模型对影像进行初步降噪。为进一步提高初步降噪结果的影像质量,有效去除混合噪声中的椒盐噪声,保留更多的影像边缘细节及纹理特征。该方法第二阶段是在自适应中值滤波的基础上采用最近邻域像素加权中值替换原滤波窗口中值,对初步降噪结果进行二次处理,得到遥感影像混合噪声最终去噪结果。为验证算法的可行性和有效性,进行了遥感影像去噪实验及去噪影像边缘检测实验。分析实验结果,无论从主观视觉还是客观评价指标上进行对比,提出的方法对于遥感影像混合噪声去噪效果优于传统去噪方法,并且能够较好地保留影像边缘细节及纹理特征,获得更清晰的影像结果。
Remote sensing image denoising is of great significance for the subsequent use and research of images.The Gaussian noise and pretzel noise are common noises in images,and the current denoising algorithms generally have the disadvantages of poor denoising effect and blurred image edges after denoising for this kind of mixed noise.To address the above problems,a second-order removal method of mixed noise in remote sensing images was proposed.In the first stage of the method,dilated convolution was introduced into the DnCNN network framework to increase the receptive field of the network,so that more feature information could be extracted from remote sensing images.Meanwhile,the DropoutLayer was introduced after the deep convolution layer to build the noise reduction model to prevent the network from overfitting and simplify the training difficulty,and then the model was used to perform the preliminary noise reduction on the images.In order to further improve the image quality of the preliminary noise reduction results,the salt and pepper noise in the mixed noise was effectively removed,and more image edge details and texture features were preserved.In the second stage of the method,the original filter window median of the adaptive median filter was replaced by the nearest neighbor pixel weighted median,so that the preliminary noise reduction result was subjected to secondary processing,and the final denoising result of the mixed noise of the remote sensing image was obtained.In order to the feasibility and effectiveness of the algorithm were verified,the remote sensing image denoising experiment and the denoising image edge detection experiment were carried out in this paper.Analysis of the experimental results shows that the method proposed in this paper is better than the traditional denoising method for the mixed noise denoising effect of remote sensing images,no matter from the subjective vision or the objective evaluation index,and the edge details and texture features of the image can be better preserved,and the image results can be obtained clearly.
作者
张胜国
任超
王子彦
闫志恒
刘桃林
郭玥
张旭东
ZHANG Sheng-guo;REN Chao;WANG Zi-yan;YAN Zhi-heng;LIU Tao-lin;GUO Yue;ZHANG Xu-dong(College of Geomatics and Geoinformation,Guilin University of Technology,Guilin 541006,China;Guangxi Key Laboratory of Spatial Information and Geomatics,Guilin 541006,China)
出处
《科学技术与工程》
北大核心
2022年第30期13219-13226,共8页
Science Technology and Engineering
基金
国家自然科学基金(42064003)。
